embedded systems
Introduction to GSM
and GSM mobile RF
transceiver derivation
The GSM system works properly
only when its component parts
operate within precise limits.
Learn the key measurements
required for testing
GSM transceivers.
By Paul Kimuli
T he GSM system was specified by the European
Telecommunications Standards Institute
(ETSI at www.etsi.org). GSM has evolved into
GSM900, DCS1800 (also known as PCN) and
PCS1900 (in the USA). GSM is now a global sys-
tem for mobile communications, spanning Europe,
Asia, Africa and much of South America.
The GSM standards define a radio communica-
tions system that works properly only if each com-
ponent part operates within precise limits.
Essentially, mobiles and base stations must trans-
mit enough power with sufficient fidelity to main-
tain a call of acceptable quality, without transmit-
ting excessive power into the frequency channels
and timeslots allocated to others. Similarly,
receivers must have adequate sensitivity and selec-
tivity to acquire and demodulate a low level signal.
This document provides an overview of the key
measurements required for testing GSM trans-
ceivers. It also discusses GSM mobile performance
derivation. It is intended to help an RF designer
with no GSM RF system knowledge get up to
speed with the GSM system.
Figure 1. Visualization of the three areas of concern in the spectrum
12 www.rfdesign.com
GSM modulation
GSM900 is the original GSM system. It uses fre-
quencies in the 900 MHz band (numbered one to
124), and is designed for wide area cellular opera-
tion with maximum output powers of 1 W to 8 W
allowed for mobile applications.
GSM uses a digital modulation format called 0.3
Gaussian minimum shift keying, or 0.3 GMSK.
The 0.3 describes the bandwidth of the Gaussian
filter in relation to the bit rate.
GMSK is a special type of digital FM modulation.
Ones and zeroes are represented by shifting the RF
carrier by plus or minus 67.708 kHz. Modulation
techniques that use two frequencies to represent ones
and zeroes are called frequency shift keying (FSK). In
the case of GSM, the data rate of 270.833 kbps is cho-
sen to be exactly four times the RF frequency shift.
This has the effect of minimizing the modulation spec-
trum and improving channel efficiency. FSK modula-
tion where the bit rate is exactly four times the fre-
quency shift is called minimum shift keying (MSK). In
GSM, the modulation spectrum is further reduced by
applying a gaussian pre-modulation filter. This slows
down the rapid frequency transitions, which would
otherwise spread energy into adjacent channels.
0.3 GMSK is not phase modulation. It is the fre-
quency shift, or change of phase state, that conveys
information. GMSK can be visualised from an in-
phase and quadrature (I/Q) diagram. Without the
Gaussian filter, if a constant stream of ones is being
transmitted, MSK will effectively stay 67.708 kHz
above the carrier center frequency. If the carrier
center frequency is taken as a stationary phase ref-
erence, the 67.708 kHz signal will cause a steady
increase in phase. The phase will role 360 degrees
at a rate of 67,708 revolutions per second. In one bit
period (1/270.833 kHz), the phase will get a quarter
of the way round the I/Q diagram, or 90 degrees.
Ones are seen as a phase increase of 90 degrees.
Two ones cause a phase increase of 180 degrees,
three ones 270 degrees and so on. Zeroes cause the
same phase change in the opposite direction.
The exact phase trajectory is tightly controlled.
GSM radios need to use digital filters and I/Q or
digital FM modulators to accurately generate the
correct trajectory. The GSM specification allows no
more than 5 degrees RMS and 20 degrees peak
deviation from the ideal trajectory.
GSM transceiver measurements
GSM mobile transmitter and receiver measure-
ments originate from section 05.05.V8.12.0 (“Radio
Access Network: Radio Transmission and Reception,”
Release 1999) of the ETSI 3GPP standards.
Performance is critical in three areas; in-chan-
nel, out-of-channel, and out-of-band. An example of
a spectrum of the three areas is shown in figure 1.
In-channel measurements determine the link
quality seen by the user in question. Measurements
include phase error and mean frequency error,
mean transmitted RF carrier power, and transmit-
ted RF carrier power versus time.
Continued on page 16
June 2003

Figure 2. Test procedure to derive phase error
Out-of-channel measurements deter-
mine how much interference the user
causes other GSM users. These mea-
surements include spectrum due to
modulation and wideband noise, spec-
trum due to switching, and Tx and Rx
band spurious.
Out-of-band measurements deter-
mine how much interference the user
causes other non GSM users of the
radio spectrum, such as the military,
police, and aviation. All other spurious
(such as harmonics and wideband) are
included here.
Phase error and frequency error
Phase error is one of the parameters
used in GSM to characterize modula-
tion accuracy. Poor phase error usually
indicates problems with I/Q baseband
generators, filters, modulators, or
amplifiers in the transmitter circuitry.
Frequency error measurements indi-
cate poor synthesizer/phase lock loop
performance (such as if synthesizers
may not settle quickly enough as they
shift frequencies between transmis-
sions). In GSM systems, poor frequency
error can cause target receivers to fail
to gain lock to transmitted signals. Also
the transmitter could cause interfer-
ence with other users.
To measure phase and frequency
error, test sets can be used to sample
transmitted output of the devices under
test to capture the actual phase trajecto-
ry. This is then demodulated and the
ideal phase trajectory is derived mathe-
matically. Subtracting one from the
other gives error signals. The mean gra-
dient of these signals gives frequency
error. The variation of this signal is the
phase error and is expressed in terms of
root mean square (rms) and peak. Figure
2 demonstrates this test procedure.
Figure 3 shows a measurement of
the phase error on one transmitted
burst and how it relates to the limits
set by the GSM standard.
Mean transmitted output power
GSM systems use dynamic power
control to ensure that each link is
16
Figure 3. Example of an absolute phase error in one burst in relation to GSM
standards
maintained sufficiently with a mini-
mum of power. This allows overall sys-
tem interference to be kept to a mini-
mum, and in the case of an MS, battery
life is maximized.
Power measurements outside of
specifications usually indicate a fault in
the power amplifier circuitry, the cali-
bration tables or the power supply.
The mean output power is mea-
sured during the useful part of the
GSM burst. When performing this
measurement, the GSM test equip-
ment derives the correct timing refer-
ence by demodulating incoming sig-
nals and gating over the useful part of
the GSM burst.
The transmitters of the GSM sys-
tems must ramp up and down within
the time division multiple access
(TDMA) structure to prevent adjacent
timeslot interference. If transmitters
turn on too slowly, data at the begin-
ning of the burst might be lost, degrad-
ing link quality. If the transmitters
turn off too slowly, the user of the next
time slot in the TDMA frame will expe-
rience interference.
Therefore, transmitted RF carrier
power versus time measurements are
performed to assess the envelope of car-
rier power in the time domain against a
prescribed mask. The measurements
also check that the transmitters’ turn
offs are complete. If a transmitter fails
this measurement, it usually indicates
a problem with the units PA or power
control loop.
Adjacent channel power
As part of the out-of-channel mea-
surements, the adjacent channel
power (ACP) is defined by two mea-
surements: spectrum due to modula-
tion and wideband noise, and spec-
trum due to switching. These two mea-
surements are usually grouped togeth-
er and referred to as output RF spec-
trum (ORFS).
The modulation process in a trans-
mitter causes continuous wave carriers
to spread spectrally. The spectrum due
to modulation and wideband noise
www.rfdesign.com
measurement is used to ensure that the
modulation process does not cause
excessive spread. This would cause
interference to adjacent channel users.
To perform these measurements,
analyzers are tuned to spot frequen-
cies and time gated across part of the
modulated burst. Using this mode, the
power is measured. The analyzer is
then retuned to the next frequency or
another offset of interest. This process
continues until all offsets are mea-
sured and verified against permissible
limits. The result of these measure-
ments is a set of frequency versus
power points that define the spectrum
of the signals. However, spectral com-
ponents that result from the effect of
bursting do not appear because the
ramps are gated out.
The test limits for these measure-
ments are expressed in dBc (power
below carrier). It follows that the first
step of the measurement is to take a
reading of the center frequency to
which the transmitter is tuned.
Spectrum due to switching
GSM transmitters ramp RF power
rapidly. The transmitted RF carrier
power versus time measurements
described earlier ensure that this
process happens at the correct times
and is fast enough. However, if RF
power is ramped too quickly, undesir-
able spectral components exist in the
transmission. This measurement also
ensures that these components stay
below the acceptable level.
To perform spectrum due to switch-
ing measurements, the analyzers are
tuned to and measure multiple offset
frequencies in zero span mode with no
time gating.
Spurious measurements
The out-of-channel measurements
are necessary to ensure GSM transmit-
ters do not place energy into the incor-
rect parts of the spectrum. This would
cause interference to other users of the
spectrum. These anomalies are referred
to as spurious transmissions.
June 2003
Figure 4. Single LNA with active mixer
The spurious transmissions are mea-
sured by connecting test sets directly to
the antenna connectors of the MS. Due
to the antennas direct connection to the
test sets, these measurements are
referred to as the conducted spurious
measurements. Measurements of this
parameter include Tx/Rx band spuri-
ous, cross band spurious, and out-of-
band spurious.
The spurious measurements can be
categorized as Tx or Rx depending on
the band they inhabit. The Tx band
spurious measurements relate to spuri-
ous that fall within the 925 MHz to 960
MHz GSM Tx band. The Rx band spu-
rious measurements, however, are
measures of how much energy the
transmitters put in the Rx band (880
MHz to 915 MHz). This test ensures
that Tx spurious don’t “jam” or desensi-
tize adjacent receivers. The specifica-
tions of the measurements are based on
1 m average distance between mobiles.
For the purpose of attenuating the Tx
band signals during these measurements,
the test setups usually include Rx band
pass filters in front of the analyzer inputs.
In some countries GSM900 and
DCS1800 systems co-exist. For this
reason the ETSI 3GPP standards
require specific cross-band performance
capability. This is to ensure the GSM
transmitters place the minimum ener-
gy required into both the DCS1800 and
GSM9000 bands.
The out-of-band spurious is a series
of spectrum analyzer measurements
over a large frequency range — from
100 kHz through 12.75 GHz. The 3GPP
standards were written to include
wideband spurious limits to which an
MS must conform.
Receivers
Sensitivity is the fundamental mea-
sure of receiver performance. It speci-
fies the minimum signal level for a
specified percentage of errors in the
demodulated information. The reported
value for all receiver measurements is
bit error rate (BER) or one of the fol-
lowing a variations:
18
Figure 5. Dual LNA with active mixer
• Frame erasure rate (FER) — The
percentage of erased frames com-
pared to the total number of frames
sent during an observation period.
• Residual bit error rate (RBER) —
When frames are erased, the BER
of the remaining frames is mea-
sured. The RBER parameter
defines this measurement.
BER is a ratio of bits received erro-
neously versus total number of bits
received. It is measured as follows: The
test systems output signals carrying
known bit patterns (usually pseudo
random bit sequences or PRBS). PRBS
signals are usually labeled PNx, where
x is the number of bits being permutat-
ed in the sequence (such as PN9 = 29 - 1
or 511 bits).
During the measurements, the
receivers under test attempt to demodu-
late and decode these patterns. By
return paths (using a method known as
loop back), the receivers send the resul-
tant bits back to the test systems for
comparison. The test systems then calcu-
late the required metrics. GSM handsets
are tested using this loop-back method.
Most receivers are required to main-
tain a specified BER in the presence of
interfering signals within the channel.
For GSM, the performance is measured
as follows: Digitally modulated signals
power levels are set 20 dBs above
receiver sensitivity at the center of the
receivers’ passband. These signals are
combined with GMSK modulated inter-
ferers. Combined signals are then
injected into the antenna ports of the
receivers. The power levels of the
GMSK interfering signals are then set
to nominal levels at which the receiver
BERs must not exceed the receiver sen-
sitivity specifications. The difference in
power levels between the two signals is
the interference ratio.
Receiver blocking constitutes one of
the out-of-channel receiver tests.
Blocking tests verify correct receiver
operation in the presence of out-of-
channel signals and monitor the
receivers’ susceptibility to internally
generated spurious responses. There
www.rfdesign.com
are three key tests that define the
receivers blocking performance: spuri-
ous immunity, intermodulation immu-
nity, and adjacent channel selectivity.
Spurious immunity is the ability of
the receivers to prevent single, out-of-
channel interference signals from caus-
ing undesired in-channel responses at
the output of the receivers. Spurious
may be generated within the receivers
from power supply harmonics, system
clock harmonics or LO spurious.
Intermodulation immunity is a
measure of the receivers’ performance
in the presence of distortion products
due to intermodulation products.
These intermodulation products are
generated when more than one tone is
present at the input of the receivers.
The tones non-linearly mix to form
third-order intermodulation products.
The products of concern lie within the
receivers’ passband.
Adjacent channel selectivity is a
measure of the receivers’ ability to
process the desired modulated signals
in the presence of strong signals in the
adjacent channels. Alternate channel
selectivity is a similar test in which the
interfering signals are two RF channels
away from the receivers’ passband.
GSM mobile RF transceiver
derivation
Receiver sensitivity is related to
receiver noise figure according to:
Where the receivers’ bandwidth (180
kHz for GSM), is the baseband signal-
to-noise-ratio and the RF and broad-
band implementation gain.
The GSM standard specifies a mini-
mum -102 dBm sensitivity requirement.
Given a worse-case baseband ratio of 9
dB and a 2 dB implementation margin,
we calculate the noise figure as:
Given this worse-case NF, the
receivers’ designers can then investi-
June 2003
 MS Blocking
Frequency band Description
signal level
600 kHz ≤ |f-fo| < 800 kHz -43 dBm In band blocking
800 kHz ≤ |f-fo| < 1.6 MHz -43 dBm
1.6 MHz ≤ |f-fo| < 3 MHz -33 dBm
3 MHz ≤ |f-fo| -23 dBm
900 MHz to 915 MHz -5 dBm Out of band blocking
980 MHz to 12750 MHz 0 dBm
Table 1: Blocking signal levels where GSM is expected to perform

gate various front end gain and NF par-
tition options according to the equation:
Where Fi is the noise factor of the
i’th block in the partition (i = 1, 2, 3 ...).
Although the second equation shows
that the higher the gain of the first
active stage, the lower the NF of the
system would be, the receivers’ design-
ers need to ensure that the first active
stage does not compress the subsequent
stages. This would degrade receiver lin-
earity. This shows that system sensitiv-
ity is a compromise between receiver
NF (dominated by choice of front end
components) and receiver linearity.
The receivers’ front end low noise
amplifier (LNA) options typically inves-
tigated are shown in figures 4 and 5.
The main benefit of the dual LNA
option — when compared with the
first option — is that the individual
LNA noise figure and gain require-
ments are significantly relaxed. With
the single LNA option, the front end
LNA would need to be tightly specified
to achieve the same system NF as the
dual LNA option.
The main disadvantage of the dual
LNA option typically is increased cost
and (potentially) the extra supply cur-
rent required due to addition of a sec-
ond LNA.
Rx blocking analysis
Table 1 shows the blocking signal
levels with which a GSM mobile is
expected to perform without drop-
ping a call. GSM receiver designers
specify the receive strips’ compres-
sion points based on the above listed
in-band blocking signal level specifi-
cations, and uses the out-of-band
blocking signal levels to define the
filter rejection specifications to avoid
signal path compression.
For example, in-band blocking at 3
MHz offset (such as -23 dBm) sets the
compression point required for the
front end. Assuming a 1 dB loss switch
and a 2.5 dB loss filter prior to a LNA
stage in the receive strip, it places a
total 3.5 dB loss prior to the LNA stage.
This means the LNA compression point
must, in the worst case, be -26.5 dBm
(such as -23 dBm to 3.5 dBm).

Offset frequency dBc/BW Derived Phase noise

±200 kHz -30 kHz/30 kHz -75 dBc/Hz
±250 kHz -33 kHz/30 kHz -78 dBc/Hz
±400 kHz -60 kHz/30 kHz -105 dBc/Hz
±600 kHz - 1200 kHz -60 kHz/30 kHz -105 dBc/Hz
±1200 kHz - 1800 kHz -60 kHz/30 kHz -105 dBc/Hz
±1800 kHz - 3000 kHz -63 kHz/100 kHz -113 dBc/Hz
±3000 kHz - 6000 kHz -65 kHz/100 kHz -115 dBc/Hz
> ±6000 kHz -71 kHz/100 kHz -121 dBc/Hz
Table 2: ETSI 05.05 specifications for spectrum due to modulation

20 www.rfdesign.com June 2003

Rx intermodulation
GSM receiver intermodulation per-
formance is predominantly affected by
the front end circuitry. The intermodu-
lation performance is affected if the IF
filter chosen has good enough attenua-
tion at +/-800 kHz and +/-1600 kHz (the
offset frequencies for which this para-
meter is tested as specified by the GSM
05.05 standard).
The equation typically used to deter-
mine system IP3 requirement is:

where

is the interference signal level (-49

dBm from GSM 05.05 spec),

is the useful signal and

is the carrier-to-interference ratio for

which the receiver is designed. The
useful signal level can be determined
from the GSM sensitivity level by
adding 3 dB. So, for an 8 dB carrier
interference ratio, the minimum GSM
receiver input intercept is -20 dBm.

Conclusion
This article is meant to only be an
overview of the key measurements
required for testing GSM transceivers,
and GSM mobile performance deriva-
tion. It is intended to help an RF
designer with no GSM RF system
knowledge get a better appreciation of
GSM module specifications and how
they can affect system performance.

About the Author

Paul Kimuli is a senior RF field appli-
cations engineer for Maxim Integrated
Products Inc. (www.maxim-ic.com)
based in the United Kingdom. He has
a first class degree with Honours
in electronic and electrical engi-
neering, from the University of
Leeds. Kimuli can be reached at
pserukenya@hotmail.com.

RF Design www.rfdesign.com 21